Sardinia, the second largest Mediterranean island with a current population of 1.65 million, has long been considered a privileged observatory for the study of medical conditions. Islands are often intriguing for epidemiologists and geneticists because they may represent geographically isolated pockets with unique populations. Just to give an example regarding the psychiatric field, the series of studies started in the 1930s by Erik Strömgren in the island of Bornholm, Denmark are among the most frequently cited epidemiological genetic works in the history of psychiatry [1]. In his doctoral thesis, Strömgren provided a complete census of psychoses existing in Bornholm in 1935, and the census was repeated half a century later [2].

The peculiar epidemiology of mood disorders and suicide across Sardinian sub-regions has long intrigued clinicians and researchers. The first attempts to survey the incidence of mental disorders in Sardinia municipalities were published in the 1960s [3, 4]. In the 1970s, Athanasios Koukopoulos noted that a strikingly high number of his patients with bipolar disorder came from one small village in Sardinia. After first consulting with Strömgren, Koukopoulos tried to undertake an epidemiological study there, but the progress in this unfunded community survey was limited [5].

Our group started to study bipolar and related disorders in Sardinia in the 1980s. The high prevalence of G6PD deficiency in Sardinia prompted us to survey G6PD activity in our population of patients with the initial aim of identifying pedigrees for an X-linkage study [6]. Unexpectedly, we found an excess of G6PD deficiency in the subgroups of patients with bipolar spectrum disorders [7-10]. We suggested that the uneven distribution of G6PD deficiency throughout Sardinia, depending on the once malaria endemicity, may explain in part why bipolar syndromes appear to be enriched in some areas, as already suggested by the two 1960s surveys of psychiatric hospitalizations [3, 4].

In the meantime, we surveyed also heterozygous β-thalassemia as a phenotypic marker to be used in a linkage study with chromosome 11p15, which was popular in psychiatric genetics at that time. Unexpectedly, similarly to G6PD deficiency, we found excessive proportions of heterozygous β-thalassemia among subgroups of patients with bipolar spectrum disorders, especially those with psychotic symptoms [7, 11, 12]. One possible explanation of such disproportions is that the two genetic conditions typical of Mediterranean areas have pleiotropic effects and play a role in psychiatric syndromes, even though a bias due to geographical stratification cannot be ruled out. With the advent of DNA molecular markers, our group undertook several genetic association studies of candidate loci for bipolar illness, but results were not substantial [13-22].

Over the last decades, we have also investigated clinical characteristics of bipolar spectrum disorders and suicide behavior within families of Sardinian patients [23, 24]. More recently, we focused again on the geographical distribution of mental illness across Sardinian sub-regions. For example, we found higher suicide mortality rates in municipalities whose lithium content in stream sediments is lower (submitted for publication).

Given the known relationship between suicide and major psychiatric illness, the aim of the present study was to test whether the geographical distribution of suicides committed in Sardinia over the last three decades may be comparable with the geographical origin of patients hospitalized in the two Sardinian mental hospitals up to half a century ago.

The availability of data on past hospitalizations and recent suicides divided by municipality has allowed the principal interesting observation from the present study that major psychoses and suicide have had similar distributions within Sardinian sub-regions across one century. A secondary observation was that the correlation with altitude and the rural/urban ratio were similar between hospitalizations and suicides.

We must underscore that both mental illness and suicide are complex phenomena with many underlying factors. Their shared geographical distribution, as found in the present study, may depend on sociocultural, genetic and/or environmental aspects, but it can provide an initial step aiming at identifying new potential tiles in the Sardinian puzzle. Recent studies have confirmed that isolated populations can provide interesting hints in the psychiatric field, as was for example the case of the traditional Amish community [25]. The Amish Study of Major Affective Disorder (ASMAD), initiated in the 1970s has tracked several multi-generation pedigrees with high prevalence of bipolar spectrum disorders [26]. All suicides for a 100-year period (1880 to 1980) were ascertained [27]. The majority (92%) of suicide cases were diagnosed with a major affective disorder and were situated in multigenerational families with heavy loading for bipolar, unipolar, and other affective-spectrum illnesses.

The population originating from some Sardinian sub-regions can be considered as isolated as the Amish community, and may also provide interesting data. Interestingly, a recent study of the Amish families combining microsatellite and high-density single nucleotide polymorphism (SNP) genotypes with whole-genome sequence data implicated dozens of rare alleles that may interact to determine risk for bipolar disorder [28]. Other data from the Amish study suggest potential roles of functional variation of neuronal potassium channels or copy number variants which may be contributing factors in the phenotypic presentation and heterogeneity of mental illness [29, 30]. Thus, it has become evident, as also suggested from other genetic studies that single genetic variants may result in a wide range of psychopathological presentations [31-33]. The latter observation challenges the validity of current nosographic classifications and may even be consistent with the results from the Sardinian survey of hospitalizations showing similar distributions between schizophrenia and mood disorders.

The incidence rates of hospitalizations for schizophrenia appear at least two times greater than those of bipolar disorders, that is at odds with current nosology, but it must be noted that the presence of psychotic symptoms in patients with affective episodes often resulted in a diagnosis of schizophrenia up to the advent of current criteria (at least up to 1980, when DSM-III began to be used also in Italy) and to the introduction of lithium therapy in the 1970s.

The comparable geographical distribution between hospitalizations and suicide rates found in the present study is consistent with the evidence that mental illness is present among >90% of suicides [34, 35]. The lifetime risk of suicide is estimated to be 4% in patients with mood disorders [36], 8% in people with bipolar disorder [37, 38], and 5% in people with schizophrenia [39].

To date, the interpretation of the reasons of the variable geographical distribution of mental illness and suicides have largely been influenced by the perspective of the XIX century French pioneer sociologist Émile Durkheim. For example, in his book “Suicide” published in 1897 [40], Durkheim focused on family and social factors, including marital status, religion, employment, and social integration. One of the principal issues has been the rural/urban ratio in suicide rates.

In the present study, we found that both hospitalizations and suicides were more incident in small villages. Similar patterns are being observed in recent studies of other populations [41, 42].

Another interesting aspect shared with studies from other countries is the positive correlation between suicide and altitude above the sea level, that has been attributed to the effects of metabolic stress associated with mild hypoxia in individuals with mood disorders [43-45].

Besides the risk factors shared with other populations, our hypothesis is that some genetic and/or environmental peculiarities may contribute to the geographical distribution of mental illness and suicide in Sardinia. For example, our series of the aforementioned studies suggested that G6PD deficiency and heterozygous β-thalassemia, two hematological conditions typical of the Mediterranean area, can have pleiotropic effects, including a potential role in bipolar-spectrum disorders [7-12]. With regard to suicide, we also suggested a role of low cholesterol concentrations, which may depend on either genetic factors and/or specifically Mediterranean dietary habits [46]. More recently, we suggested that comorbidity with thyroid autoimmunity, which is very prevalent in Sardinia [47] may be associated with psychiatric manifestations including suicidality [48, 49]. With particular regard to the geographic distribution across Sardinian sub-regions, we would like to mention our recent finding that the concentration of lithium in stream sediments correlates inversely with suicide mortality (manuscript in preparation). A protective effect of naturally occurring lithium against suicide has already been suggested by research groups from several countries studying lithium content in drinking water (for review see [50]). The relationship between lithium content in drinking water and suicide may also have complex interactions with altitude [51], which is was in turn associated with hospitalization and suicide rates in the present study.

The present study confirms prior suggestions that there is a peculiar geographical distribution of major psychoses and suicide in Sardinia. The two phenomena appear to have been correlated for as long as one century. These data can be the base for further studies aiming at unraveling the complex pathogenesis of major psychoses and suicide. One potential approach, similar to that already used in the Amish study [29], would be to search for candidate genetic variants carried by affected members of families originating from specific Sardinian isolates.